Swarm-MAPET: 16-axis Byzantine consensus for post-quantum receipts

Swarm-MAPET (Multi-Axis Physical Entropy Tabulation, Swarm variant) is Hive's 16-axis Byzantine fault-tolerant consensus protocol. It pulls physical entropy from heterogeneous validators and aggregates it into the seed of a cosmic-tier post-quantum receipt. This page is the protocol reference.

Why physical entropy

Algorithmic randomness — a CSPRNG seeded by /dev/urandom — is sufficient for ML-DSA-65 signing in isolation. But it is not sufficient for auditable receipt issuance, because the auditor cannot replay an algorithmic seed. They have to trust the issuer's claim about what the seed was at issuance time.

Physical entropy fixes the audit gap. If the entropy source is an external, recordable physical phenomenon — RF noise, photodiode jitter, accelerometer fluctuation — the auditor can replay the validator timestamps and verify that the seed was consistent with the recorded physical signal. The seed becomes provenance-bearing.

The 16 axes

Swarm-MAPET aggregates entropy across sixteen orthogonal physical axes. Each axis is a distinct measurement modality and each is run by a separate validator with a separate hardware vendor where possible:

The 6-axis Wave-Lattice profile uses axes 1, 3, 5, 7, 9, 11. The 16-axis RogueWave-Lattice profile uses all sixteen.

Byzantine fault tolerance

Swarm-MAPET tolerates up to ⌊(n−1)/3⌋ Byzantine validators where n is the active axis count. For the 16-axis configuration that is up to 5 corrupt validators. The protocol uses a leader-rotating BFT round with the following gates:

• Validator deduplication — each validator counts at most once per round.
• Timestamp freshness — validator timestamps must fall within a 5-minute window with at most 30 seconds skew.
• Signature surface gate — every validator vote is signed with ML-DSA-ready keys.
• Epoch enforcement — epoch=0 fails closed; the protocol does not finalize a vacuous round.

The leader proposes a candidate seed (the XOR-aggregate of axis readings); validators vote; finalization requires a 2/3+ supermajority of distinct, fresh, signed votes.

Quorum and timing

The cosmic profile finalizes a seed in approximately 250 ms under nominal load. The swarm-tier (federated mesh, multi-node) profile takes longer — typically 600–900 ms — because it requires a cross-node quorum on top of the single-node 16-axis vote. That extra latency is the cost of cross-node Byzantine resistance.

How the seed enters the receipt

Once the BFT round finalizes, the aggregated seed is written into the mapet_seed field of the receipt envelope. The receipt also records:

• The set of validator DIDs that participated.
• The leader DID.
• The round timestamp.
• A SHA-256 commitment to the per-axis readings (so they can be replayed if the validator publishes them).

The auditor can later request the per-axis readings from each validator and verify that the commitment matches.

The three parked surfaces

Three deployment surfaces are reserved for Swarm-MAPET and announced publicly here for the record:

• Embed-in-phones — ARM TrustZone / Secure Element implementation. Mobile-side validator participation.
• RogueWave-Lattice SoC — purpose-built silicon, target $15 BOM at 100K volume (Build #22 in our internal roadmap).
• Industrial / sovereign / orbital — hardened deployments for industrial sensing, sovereign-grade nation-state deployments, and orbital validators.

Cosmic and Swarm pricing